Fig 1.
Single-nuclei profiling of human heart.
(a, b) Distribution of age groups (a) and different conditions (b) among samples in the integrated data. (c) Annotation of cell types in the integrated data. (d) Expressions of marker genes derived from distinct cell types among cell clusters. (e) Statistics on the abundance of each cell type. (f) Condition preference of each cell type evaluated by the Ro/e index. (g) The difference in the proportion of neurons between the heart failure group and the normal control group. Wilcoxon rank-sum test was used. LEC, lymphatic endothelial cell.
Fig 2.
Identification of neuronal subsets associated with heart failure.
(a) The heatmap displaying the top five differentially expressed genes in each neuronal cluster. (b) Definition of neuronal subsets. (c) Differences in the frequency of each neuronal subset between heart failure and healthy groups. (d) Comparisons upon the percentage of neuronal subsets between heart failure patients and healthy donors. (e) Overlapping of differential genes in neuronal subsets between our generated data and SCP1303 dataset. Hypergeometric test was used. (f) The enrichment scores of N1-XKR4, N2-OGFRL1, and N4-ALK marker genes in the SCP1303 dataset. (g) The expression of marker genes from N4-ALK, N1-XKR4, and N2-OGFRL1 over neuronal clusters in the dataset of SCP1303. (h) Left Changes in the frequency of neuronal subsets before and after receiving LVAD implantation in patients who did not respond to the treatment. Center Changes in the frequency of neuronal subsets in response to LVAD before and after receiving the treatment. Right Changes in the frequency of neuronal subsets at the statuses of responsive and unresponsive after receiving LVAD implantation. (i) The differential expression of LRRC4C in response to LVAD before and after receiving the implantation. *P < 0.05, **P < 0.01, ***P < 0.001, ns, no significance. In (d) and (i), P value was calculated by Wilcoxon rank-sum test. R or NR state, LVAD responsive or unresponsive status; Post State, after receiving LVAD implantation; R or NR pre, before LVAD implantation for treating individuals with or without response; R or NR post, after LVAD implantation for individuals with or without response.
Fig 3.
The transcriptional heterogeneity of neuronal subsets.
(a) GO enrichment analysis of differentially expressed genes in N1-XKR4, N2-OGFRL1, and N4-ALK, respectively. The q value represented P value adjusted by Benjamini & Hochberg (BH). (b) Pseudo time analysis of reshaping neuronal cells over time. (c) Different cellular states divided by four branch nodes. (d) Evaluation of the differentiation ability on each neuronal cell. The color ranging from blue to red represented an increase in differentiation ability. (e) The composition of cellular states in heart failure and healthy conditions respectively. (f) Left Molecular changes along the trajectory conducted by branch three. Right GO enrichment analysis on each gene cluster with showing the top three pathways.
Fig 4.
Cellular crosstalk between neurons and fibroblasts.
(a, b) Differences in the strength of interactions between patients with DCM and healthy donors (a) and patients with HCM and healthy donors (b). The red arrow pointed out the cell types of N4-ALK on X-axis and fibroblasts in Y-axis. Black box marked the differential interaction between these two cell types. (c) Pearson correlation between the percentage of fibroblasts and the percentage of each neuronal subset in patients with heart failure. (d) Dot plot showing the communication between fibroblasts and neuronal subsets mediated by specific ligand-receptor pairs. (e) Differential expression of PTPRZ1 and ALK in neurons between the heart failure group and the healthy group. (f) Differential expression of PTPRZ1 before and after receiving implantation in the LVAD responsive group. In (e) and (f), Wilcoxon rank-sum test was used.
Fig 5.
Identifying the transcriptional regulatory mechanism of neuronal cell in heart failure.
(a) Five regulons with higher transcriptional activity in N4-ALK cells. (b) Differential expression of transcription factors ATF1, NRL, PAX3, RFX1, and RXRG in different cell types. (c) Transcriptional networks consisting of the five transcription factors and their top ten targets. (d) The density plot showing the enriched regions of RXRG regulon activity on the neuronal map. (e) The expression of RXRG across neuronal subsets. (f) Differential expression of RXRG in heart tissue between heart failure group and healthy group both in GSE165303 and GSE57345. (g) Pearson correlation between the expression of RXRG and PTPRZ1 both in GSE165303 and GSE57345. (h) Differential expression of RXRG in response to LVAD before and after receiving LVAD implantation. (i) Differential expression of RXRG in cardiac neurons with or without response to LVAD after receiving implantation. (j) The crosstalk between cardiac neurons and fibroblasts mediated by overexpression of RXRG in heart failure. Images provided by Servier Medical Art (https://smart.servier.com), licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). In (f), (h), and (i), Wilcoxon rank-sum test was used.
Fig 6.
Analysis of the molecular mechanism mediated by the newly generated neuronal subsets.
(a) Label transfer learning analysis on the newly generated neuronal map included different etiologies. (b) Differences of the proportion in neuronal subsets across different datasets. (c) The differential expression of PTPRZ1 and ALK between normal heart and diseased heart induced by various etiologies. Wilcoxon rank-sum test was used. (d) Cell-cell communications of PTN signaling pathway. (e) Pairs of ligand and receptor in the interactions between fibroblasts and neurons in the failing heart.
Fig 7.
Construction of a predictive model for heart failure using N4-ALK associated features.
(a) Identification of co-expressed gene modules within neurons. (b) The scores of four gene modules in ten subsets of neurons. (c) Overlapping genes of “yellow” module and genes detected in data sets GSE141910, GSE165303, GSE145154, and GSE57345. (d) Heat map displaying the AUC score of each heart failure prediction model, and bar plot illustrating its average AUC score across four datasets. (e) The differential enrichment scores of 66 genes between the heart failure group and the healthy group in GSE141910 and GSE145154, respectively. Wilcoxon rank-sum test was used. AUC, area under the ROC curve.